Nano Banana MCP Server
# nano-banana-mcp
# Nano Banana MCP Server 🍌
A Model Context Protocol (MCP) server that connects Claude to **Nano Banana** (Google's `gemini-2.5-flash-image` AI model), allowing Claude to seamlessly generate and edit images directly in your chat.
## Features
- **`generate_image`**: Prompt Claude to generate high-resolution images. Features adjustable aspect ratios (`1:1`, `16:9`, `9:16`, `3:4`, `4:3`).
- Saves generated images automatically to your local machine and returns the exact file path to Claude.
## Prerequisites
- Node.js (v18 or higher)
- A Google Gemini API Key (from [Google AI Studio](https://aistudio.google.com/))
- Claude Desktop App
## Setup & Installation
1. **Clone the repository:**
```bash
git clone [https://github.com/yourusername/nano-banana-mcp.git](https://github.com/yourusername/nano-banana-mcp.git)
cd nano-banana-mcp
Install dependencies and build:
Bash
npm install
npm run build
Configure Claude Desktop:
Open your Claude Desktop configuration file:
Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Add the MCP server to your config (replace /path/to/repo with the actual absolute path to this folder, and insert your API key):
JSON
{
"mcpServers": {
"nano-banana": {
"command": "node",
"args": ["/path/to/repo/nano-banana-mcp/build/index.js"],
"env": {
"GEMINI_API_KEY": "your_actual_api_key_here"
}
}
}
}
Restart Claude Desktop:
Fully quit and restart the Claude Desktop application. You will now see the generate_image tool available (look for the hammer icon ⚒️ in your chat).
Usage Example
Simply ask Claude:
"Hey Claude, use Nano Banana to generate a 16:9 image of a futuristic cyberpunk city covered in neon lights."
---
### 3. Finalizing for GitHub
1. Ensure you compile the code before pushing or running:
```bash
npm run build
Create a .gitignore file so you don't commit your build files or node modules:
Plaintext
node_modules/
build/
.env
Initialize Git, commit, and push:
Bash
git init
git add .
git commit -m "Initial commit: Nano Banana MCP server"
git branch -M main
git remote add origin https://github.com/YOUR_USERNAME/nano-banana-mcp.git
git push -u origin main
TDQS
Scored across 1 tool
With only a single tool in the set, there is no possibility of confusion or misselection. The tool's purpose (image generation via a specific model) is unmistakable.
The lone tool follows a clear verb_noun snake_case convention (generate_image). There is nothing to be inconsistent with, so the naming is fully predictable.
One tool is a thin surface even for a focused image-generation wrapper; common needs like editing, variation, or configuration are not expressible. It is borderline rather than mismatched, since the server targets a single model.
Generation is covered, but an image-model server typically also supports editing/inpainting, variations, or parameter control (size, count, seed), none of which exist here. Agents can work with the single entry point but will hit dead ends for anything beyond plain generation.